By Hamza Faisal

Canva has released its first image generation model, Canva Craft, built in-house by its research team to compete directly with the image models from Google, OpenAI, ByteDance, and Meta. The model is live now in the company's Dream Lab tab for all users, marking the first public release from Canva Research.

Craft is designed to turn a text prompt into an image of up to 2K resolution in any of more than 30 aspect ratios, with support for custom sizes. Canva says the model can render legible text inside images in 55 languages, and that users can make precise edits to one element without changing the rest of the picture.

A model that plans before it paints

Most image models turn a prompt straight into pixels. Craft splits the job in two. In the first stage, a planning model interprets the prompt and writes out a detailed plan: the spatial relationships of elements, their relative scale, visual hierarchy, and the text that needs to appear in the image. In the second stage, a separate rendering model turns that plan into the final picture.

Canva says this plan-first approach produces images with accurate object details and readable text, a persistent weak spot for rival generators. The planning structure also means edits stay local — change a headline's wording or move an object, and the surrounding layout remains untouched.

Rather than photographs, Canva trained Craft on its own library of human-designed templates: posters, social posts, presentations, infographics, menus, and typography-heavy layouts. Because the training data already carries design structure, images generated with Craft can be converted into layered, editable designs through Canva Magic, the company's image-to-design transformer, which rebuilds flat images into separate text boxes, objects, and backgrounds.

Not just a product launch — a cost play

Until now, Canva has leaned on other companies' image models. Its Dream Lab, launched in October 2024, ran on Phoenix, the model from Leonardo.ai, the Sydney startup Canva acquired that same year. Researchers from Leonardo contributed to Craft, the company said.

Building its own model changes the economics. Canva has used a mix of external image models across its platform, and owning the stack cuts third-party API costs while giving the company tighter control over quality and speed. The company is already developing two cheaper variants, Craftfast and Craft-ultrafast, for release soon. The launch follows a reported cut to Canva's 2026 growth forecast — external analysts reduced its valuation from $42 billion to $35 billion even as the company was on track for $4 billion in revenue last year.

Fourth place, second-best lab

Canva reports that Craft debuted fourth among image models on DesignArena, the independent leaderboard that ranks models through community comparisons. That ranking sits ahead of entries from Google, Microsoft, Meta, xAI, and ByteDance — which the company frames as making Canva the second-best AI lab for image generation in the world, behind only OpenAI.

Craft is available to all users through the Dream Lab tab and will also be offered through the Canva API. The company plans to support image generation in third-party chatbots such as ChatGPT and Gemini through Canva's MCP connector.

For a platform used by more than 250 million people, the move turns Canva from an AI customer into an AI lab. The question now is whether designers will treat Craft as just another generator, or as the start of a genuine alternative to the incumbents.

— Hamza Faisal